ABSTRACT

Many researchers have worked on the Software fault forecast model because the software fault forecast is very important in software development projects. In terms of the Software fault forecast model, earlier researchers have examined defective datasets models with the help of metrics and classification methods. Classification is assuming an exceptionally major job in the software fault forecast model, which is an important issue in data mining. The machine learning system as a finding way for the information securing or information extraction issue has examined it widely. The contribution to a classifier is a training data set of precedents, every one of which is labeled with a class name. Classification separates data tests into target classes. Software modules are categorized as defected models or not defected models by classification draws near. In Classification, class categories are known thus it is a supervised learning approach. In our research, software fault forecast datasets models are examined with the help of tree vector machine classification. Our proposed model is a tree vector machine, which is used for increasing the positive accuracy and efficiency of the software fault forecast model. We have used multiple tree classifiers for getting more accurate results and compare them with each other. During the analysis of the experiments, j48, random forest and random tree have increased their performance in accuracy as well as efficiency. However, the performance of REP Tree, Hoeffding Tree and Decision Stump is not so good at all measure rates. The experiment's analysis results showed that not every tree classifier could be good in all in measure unit.

Keywords: - Software, Fault Forecast, Classification, Defect prone, Support Vector Machine, J48, Random Tree